|The paper is as previously very interesting and has greatly improved through the revision:|
Some remaining comments:
Table 2 and 3: I'd strongly suggest to include GLEAM validation metrics.
Why is no validation of GLEAM included? Surely this is vital as GLEAM is very often used for global studies and the paper should provide insights in how, when or where the merged product surpasses GLEAM. Note that when validating E products it is often standard-practise to exclude days with strong precipitation.
If GLEAM outperforms the others for a given pixel, can the merged product actually achieve the same or better performance than the GLEAM reference data? I'm just wondering, some thoughts and a sentence or two would be helpful.
Furthermore I'd expect a few sentences on the exact input data GLEAM version a relies on based on the respective GLEAM paper. Net radiation as correctly stated is based on ERA-Interim and GLEAM is actually extremely sensitive to net radiation.
Concerning these sentences in the rebuttal letter/or text:
Please reword/revise parts in the updated manuscript that reflect on the ideas below.
"GLEAM is not a traditional terrestrial model ..." Please reword this, it's not clear what a 'traditional' terrestrial model is or why GLEAM should be any different. Perhaps argue that GLEAM is specifically designed to estimate evaporation whereas the other 'big' models are required to simulate a higher number of variables decently (This is a spontaneous idea, please check carefully with the literature). GLEAM is not part of a larger Earth System model with an atmospheric/sea ice component etc. Perhaps that's more of a difference too?
"MERRA2 and ERA5 are based on brightness temperatures that are assimilated into their atmospheric models and only indirectly impact the land states." --> A lot more than brightness temperatures are assimilated, e.g. IR radiances, air temperature measurements from aircraft etc. etc. etc.
"It is expected that GLEAM’s over-reliance on observations states would serve as some
sort of benchmark to estimate the weights of the model-based products. Thus, the goal
is not based on a superior skill of GLEAM but its added value due to its uniqueness
relative to the model-based products, which we believe, does have merits"
One might argue that GLEAM is more directly linked to satellite input but it is no less a model than the other products. The reanalysis products incorporate many many more observations than GLEAM does. It is a rather simple model (in a good way) focusing on soil moisture and evaporation (and computes some more variables required for E and soil moisture).
"GLEAM is not a traditional terrestrial model as found in ERA5, MERRA2 and GLDAS" See above, not sure about traditional.
"GLEAM (Miralles et al., 2011a) (Global Land-Surface Evaporation: The Amsterdam Methodology) is derived from the inversion of multi-source remote sensing data, meteorological reanalysis data and the improved Priestley-Taylor (P-T) formula"
Surprisingly it is sometimes stated that GLEAM is an inversion or retrieval method but in my view it isn't. Inversion in my understanding is based on forward simulations of something observable from satellite, e.g. brightness temperatures, radiances etc. These forward simulations are based on a model (radiative-transfer) with multiple geophysical input variables. Minimising the difference between forward simulations and a satellite observation by assuming certain geophysical conditions is a retrieval based on inversion. GLEAM does no such thing.
GLEAM is a simple land surface model focusing on the estimation of evaporation and soil moisture. It's a traditional top-down approach with a model being fed with atmospheric input and land surface conditions (e.g. vegetation phenology). The estimation of evaporation is based around the Priestley-Taylor formula.
I'm not sure about the other two evaporation products, I would assume they are also specific models and not inversion schemes at all but please check.
"Thank you very much for your comment. Indeed, GLEAM is not the only one that contains soil moisture, however, GLEAM is the only product that uses satellite retrieved soil moisture to drive the model."
Satellite retrieved soil moisture does not drive GLEAM. GLEAM computes soil moisture at different levels based on soil properties, precipitation input etc. very similarly to the other models (similar in principle, not the exact formulas). Satellite retrievals are assimilated with a very simple Newtonian Nudging scheme slightly correcting the modelled soil moisture. The impact of this assimilation is however mostly quite low. Therefore you can give equal credit to the other models with their repsective soil water modules.
L43: I suppose SiF can be used for E although data quality is still not great (I'm no expert on this).
L47: satellite inversion is incorrect, definitely for GLEAM.
L117: "GLEAM is not a traditional terrestrial model as found in ERA5, MERRA2 and GLDAS" See above, I don't understand what is meant by this.
I think it's still missing a clearer justification of using GLEAM (and the validation of GLEAM itself compared to the merged product and other individual ones).
L120: Monthly data ... for what purpose is this monthly data used?
L227: Is GLDAS a reanalysis? If yes, okay.